#Meta卖算力引发存储股大跌 Meta "selling computing power" shakes the AI market: oversupply or excessive panic?

A piece of news about Meta has sparked a fierce debate about the future of AI infrastructure in the global capital market. Currently, Meta is planning to offer AI computing services to external customers and is considering leasing out some of its idle AI computing power to improve the utilization of its AI infrastructure. This means that the internet giant, which previously relied mainly on its advertising business to digest AI investments, is beginning to attempt to directly convert its massive GPU clusters into sellable cloud services, seeking new return paths for its persistently high AI capital expenditures. The market immediately gave contrasting valuations.
On July 1 local time, Meta's stock price surged 8.8%, adding approximately $127 billion to its market cap in a single day. On the other hand, the computing power industry chain, seen as a beneficiary of AI infrastructure, collectively came under pressure. CoreWeave and Nebius plummeted 13.92% and 17.01%, respectively, while storage chip stocks like SanDisk and Micron Technology fell over 10%. The sentiment then spread to other markets. On July 2, South Korea's Samsung Electronics and SK Hynix dropped 9.06% and 14.57%, respectively. AI industry chains such as A-share storage chips, optical modules, and PCBs also experienced widespread corrections, with GigaDevice hitting the daily limit down, and multiple stocks including Zhongji Innolight, Montage Technology, and Longsys falling over 10%.
Compared to stock price fluctuations, the bigger shock came from market sentiment. Over the past two years, driven by generative AI, global tech giants have launched an unprecedented capital expenditure race. From Meta, Microsoft, and Google to Amazon, OpenAI, and xAI, the competition in large models has driven the continuous expansion of global GPUs, HBM, high-speed networks, and data centers, making AI infrastructure one of the most important investment themes in the global capital market.
Now, Meta's attempt to lease computing power externally is seen by investors as a signal: if even one of the world's largest GPU buyers starts seeking to sell "surplus" computing power, does it mean that AI infrastructure construction has reached an inflection point? Is the continuously high demand for computing power over the past two years starting to show signs of oversupply risk?
Why Meta sparked the computing power oversupply debate Meta's move is essentially a business model adjustment. In recent years, Meta has been purchasing AI chips and building data centers on a large scale. Its AI investments have mainly relied on indirect monetization through the advertising business, improving ad revenue by enhancing ad recommendation effectiveness. Financial reports show that in the fourth quarter of 2025, advertising revenue still accounted for over 90% of Meta's total revenue. Compared to its advertising business, Meta's competition in large models has not gone smoothly. In terms of model development, Meta has been continuously adjusting its AI strategy over the past year, undergoing multiple rounds of organizational restructuring. Under competitive pressure from Google Gemini, OpenAI, and others, Meta has begun exploring more ways to monetize its AI assets. Selling or leasing computing power has thus become a new option.
In fact, Meta is not the first company to do this. Previously, Musk's xAI also leased part of the computing power of its Colossus supercomputing cluster to customers like Anthropic and Google to improve GPU utilization. But market concerns have followed. Many investors believe that if leading internet companies start leasing computing power, it means that many of the GPUs they previously purchased may already be idle, and the "continuous expansion" logic that the AI industry chain previously relied on may change. Therefore, AI hardware sectors such as HBM, storage, and optical modules were the first to be sold off.
However, several industry insiders interviewed did not agree with the judgment of "full-scale computing power oversupply." A relevant person in charge of the marketing department of Jiuzhang Cloud Extreme told reporters that the current problem is not absolute oversupply, but "structural mismatch." "If oversupply refers to low-end general-purpose computing power and AI computing centers lacking application scenarios, then there is indeed a partial oversupply issue. But if it refers to high-end intelligent computing power that can support large model training and inference, and operational computing power that can be scheduled on demand, then it is still far from sufficient." Citing industry data, the source said that the effective utilization rate of current AI computing clusters averages less than 20%, but the gap in high-end computing power supporting large model training is about 40%. Currently, training computing power in the industry is in a state of supply shortage.
Ablikim Ablimit, Vice President of Lenovo Group and Chief Strategy Officer of Lenovo China, also told reporters that whether in China or overseas, the long-term demand for AI computing power still has significant room for growth, and there is currently no oversupply of computing power. He believes that truly mature C-end AI products are still limited, while B-end enterprise-level AI commercialization is just beginning. The demand release brought by previous productivity revolutions mostly came from the industrial side, not the consumer side. Therefore, future AI computing power demand still has great growth potential.
A Lenovo insider previously revealed that the company's AI servers are still in a state of supply shortage, with pending delivery orders of about 150 billion yuan. Due to the tight supply of core chips such as GPUs, some orders still need to wait in line for delivery. An insider from Unisplendour Corporation also believes that it is too early to discuss "oversupply." "The real problem is not whether there is computing power, but when AI applications can form a commercial closed loop and use application revenue to cover infrastructure investment."
Slowing capital expenditure does not mean the industry has peaked In the past two years, the AI industry competed on "who invests more," but currently, the focus of the capital market has gradually shifted from the scale of capital expenditure itself to investment efficiency.
Chen Jun, Deputy General Manager and Chief Analyst at Sigmaintell, told reporters that the closed loop of computing power investment is an indispensable part of the entire AI infrastructure investment closed loop and is also a core issue that global cloud vendors need to solve in the coming period. He predicts that in the future, AI infrastructure investment will gradually transition from extensive expansion to an efficiency improvement phase. Top cloud vendors will be more cautious when purchasing hardware such as storage, but this will not change the long-term investment direction.
Sigmaintell predicts that global AI infrastructure investment will still maintain double-digit growth from 2024 to 2028. Among them, investment in 2026 is still expected to grow 51% year-over-year. Although this is a decline from the 104% growth rate in 2025, it is still in a high-speed expansion phase. In 2027 and 2028, growth is expected to be 15% and 11%, respectively. This means that AI investment is transitioning from explosive growth to a more sustainable development stage, rather than entering a decline cycle. This trend is also reflected in the latest moves of global tech giants. Just as the market worries about AI capital expenditure peaking, South Korea's Samsung Electronics and SK Hynix recently announced plans to invest about 4,755 trillion won in South Korea in the future, ramping up AI-related businesses such as semiconductors, physical AI, and data centers. Chinese internet companies like Alibaba and Tencent have also recently reaffirmed their commitment to continue expanding AI capital expenditure, with domestic computing power deployment progressing steadily.
In the view of many brokerages, Meta's leasing of computing power is essentially optimizing capital returns, not cutting investment. Everbright Securities stated that Meta's trial of external computing power leasing is an optimization of the structure of computing power capital expenditure, with short-term valuation facing gaming pressure. The underlying logic of Meta's entry into cloud computing is to create new monetization channels for its massive AI infrastructure investment. The core issue previously questioned by the market was that Meta's AI capital expenditure was too large, and when it would be converted into profits. The development of its cloud business transforms Meta's GPU clusters from "pure cost burdens" into "revenue-generating assets." Monetizing idle computing power externally directly dilutes depreciation and operation costs. This move will effectively boost the market's expectations for its future cash flow. "Computing power leasing is a resource optimization measure that does not change medium- to long-term expansion plans, but expansion expectations may converge in stages. According to Zuckerberg's remarks at the shareholders' meeting, the company has the ability to lease some computing resources externally, indicating that the ability to monetize externally has itself become a cash flow guarantee supporting Meta's continued expansion of AI investment." Everbright Securities believes that the business model of the current AI computing power leasing track is receiving the strongest endorsement. Meta's direct involvement indirectly verifies the current high returns: referring to xAI's leasing of computing power to Anthropic, with a monthly rent of $1.25 billion, the implied ROI suggests a payback period of two years. Open Router's latest data from June shows that from June 21 to June 28, the global weekly token volume reached a new high of 46.7 trillion, a 0.2% increase month-over-month. The current situation of supply shortage remains unchanged.
Tianfeng Securities believes that Meta's move into AI cloud does not mean acknowledging a full-scale GPU oversupply. It is more likely allocating different generations of AI computing power to different economic uses. The latest GB200/GB300/Rubin and other resources will prioritize training next-generation models, while the previous generation H100/H200 will shift more toward inference and external computing power sales. Meanwhile, Google once failed to fully meet Meta's access needs for Gemini due to capacity constraints, which actually indicates that frontier models and high-quality inference capacity remain tight. "Therefore, this is not the end of AI CapEx transactions, but the evolution of the business model from pure money-burning infrastructure to chargeable platform assets. The market should not simply interpret Meta's leasing of computing power externally as the peak of AI computing power demand." Tianfeng Securities stated that for the hardware supply chain, what really needs to be observed is not "whether Meta is leasing computing power" itself, but whether the real token usage and ARR growth rates of OpenAI and Anthropic continue to rise. If the ARR and token usage of model companies continue to grow, and hyperscaler CapEx does not significantly decline, then the hardware main theme remains valid.
However, in the interview, Ablikim Ablimit also raised another issue worth noting: the current supply side of the AI industry still has relatively strong bargaining power, and the industry's supply-demand cycle has not been fully established. This judgment also aligns with the consensus among industry insiders.
He Hui, Director of Omdia China Semiconductor Analysis, told reporters that the development of large models has not stopped, and cloud vendors are also looking for business models that can be monetized. A person from Unisplendour Corporation believes that the AI industry must ultimately return to commercialization itself. "During the early stages of internet development, there was also a phase where infrastructure came first and business models lagged behind. In the future, AI must also find truly sustainable profit-generating industry scenarios to form a commercial closed loop where revenue covers investment."
From the industry's perspective, the current focus should no longer be on whether AI will continue to invest, but on when the hundreds of billions of dollars in AI infrastructure can enter a new phase of self-sustaining and self-circulating. Only when infrastructure, model capabilities, and application commercialization form a complete positive cycle can the AI industry chain truly break free from the development model that relies solely on capital investment and enter a healthier, more sustainable growth cycle.